Abstract: Systems and methods for a machine-learning driven fine-grained file access control approach are provided. According to one embodiment, a server associated with an enterprise network can obtain and store information regarding historical user behavior of users of the enterprise network by observing file access requests initiated by the users. The server receives a file access request initiated by a user, which relates to a file stored within the enterprise network in encrypted form. In response to receipt of the file access request, the server determines a risk score for the user based on multiple factors, including information regarding historical user behavior, the file access request and observed data determined based on the file access request so that based on the risk score, access to the file is permitted by returning a decryption key for the file or denied by withholding the decryption key.

Abstract: Systems and methods for a machine-learning driven fine-grained file access control approach are provided. According to one embodiment, a server associated with an enterprise network can obtain and store information regarding historical user behavior of users of the enterprise network by observing file access requests initiated by the users. The server receives a file access request initiated by a user, which relates to a file stored within the enterprise network in encrypted form. In response to receipt of the file access request, the server determines a risk score for the user based on multiple factors, including information regarding historical user behavior, the file access request and observed data determined based on the file access request so that based on the risk score, access to the file is permitted by returning a decryption key for the file or denied by withholding the decryption key.